Update Model

api/auth.py

@@ -95,8 +95,21 @@ class UserManager(IntegerIDMixin, BaseUserManager[User, int]):
             logger.info(f"Found existing OAuth account for {oauth_name}, account_id: {account_id}")
             user = existing_oauth_account.user
             if user is None:
-                logger.error(f"No user associated with OAuth account {account_id}")
-                raise ValueError("No user associated with this OAuth account")
+                logger.error(f"No user associated with OAuth account {account_id}. Creating new user.")
+                # Create new user if user is None
+                user_dict = {
+                    "email": account_email,
+                    "hashed_password": self.password_helper.hash("dummy_password"),
+                    "is_active": True,
+                    "is_verified": is_verified_by_default,
+                }
+                user = User(**user_dict)
+                self.user_db.session.add(user)
+                self.user_db.session.commit()
+                self.user_db.session.refresh(user)
+                existing_oauth_account.user_id = user.id
+                self.user_db.session.commit()
+                logger.info(f"Created new user and linked to existing OAuth account: {user.email}")
             logger.info(f"Returning existing user: {user.email}")
             return await self.on_after_login(user, request)
 
@@ -105,13 +118,25 @@ class UserManager(IntegerIDMixin, BaseUserManager[User, int]):
             statement = select(User).where(User.email == account_email)
             result = self.user_db.session.execute(statement)
             user = result.scalars().first()
-            if user is not None:
-                logger.info(f"Found existing user by email: {user.email}")
-                oauth_account.user_id = user.id
-                self.user_db.session.add(oauth_account)
+            if user is None:
+                logger.error(f"No user found for email {account_email}. Creating new user.")
+                # Create new user if not found
+                user_dict = {
+                    "email": account_email,
+                    "hashed_password": self.password_helper.hash("dummy_password"),
+                    "is_active": True,
+                    "is_verified": is_verified_by_default,
+                }
+                user = User(**user_dict)
+                self.user_db.session.add(user)
                 self.user_db.session.commit()
-                logger.info(f"Associated OAuth account with user: {user.email}")
-                return await self.on_after_login(user, request)
+                self.user_db.session.refresh(user)
+                logger.info(f"Created new user for email: {user.email}")
+            oauth_account.user_id = user.id
+            self.user_db.session.add(oauth_account)
+            self.user_db.session.commit()
+            logger.info(f"Associated OAuth account with user: {user.email}")
+            return await self.on_after_login(user, request)
 
         # Create new user
         logger.info(f"Creating new user for email: {account_email}")

api/endpoints.py

@@ -8,14 +8,14 @@ from api.models import QueryRequest, ConversationOut, ConversationCreate, UserUp
 from api.auth import current_active_user
 from api.database import get_db
 from sqlalchemy.orm import Session
-from utils.generation import request_generation, select_model
+from utils.generation import request_generation, select_model, check_model_availability
 from utils.web_search import web_search
 import io
 from openai import OpenAI
 from motor.motor_asyncio import AsyncIOMotorClient
 from datetime import datetime
 import logging
-from typing import List
+from typing import List, Optional
 
 router = APIRouter()
 logger = logging.getLogger(__name__)
@@ -31,9 +31,9 @@ if not BACKUP_HF_TOKEN:
     logger.warning("BACKUP_HF_TOKEN is not set. Fallback to secondary model will not work if primary token fails.")
 
 ROUTER_API_URL = os.getenv("ROUTER_API_URL", "https://router.huggingface.co")
-API_ENDPOINT = os.getenv("API_ENDPOINT", "https://api-inference.huggingface.co")
+API_ENDPOINT = os.getenv("API_ENDPOINT", "https://api.cerebras.ai/v1")  # تغيير الافتراضي لـ Cerebras
 FALLBACK_API_ENDPOINT = os.getenv("FALLBACK_API_ENDPOINT", "https://api-inference.huggingface.co")
-MODEL_NAME = os.getenv("MODEL_NAME", "openai/gpt-oss-120b")  # بدون :cerebras
+MODEL_NAME = os.getenv("MODEL_NAME", "openai/gpt-oss-120b")  # النموذج الرئيسي
 SECONDARY_MODEL_NAME = os.getenv("SECONDARY_MODEL_NAME", "mistralai/Mixtral-8x7B-Instruct-v0.1")
 TERTIARY_MODEL_NAME = os.getenv("TERTIARY_MODEL_NAME", "Qwen/Qwen2.5-0.5B-Instruct")
 CLIP_BASE_MODEL = os.getenv("CLIP_BASE_MODEL", "Salesforce/blip-image-captioning-large")
@@ -85,6 +85,16 @@ async def handle_session(request: Request):
         )
     return session_id
 
+# Helper function to enhance system prompt for Arabic language
+def enhance_system_prompt(system_prompt: str, message: str, user: Optional[User] = None) -> str:
+    enhanced_prompt = system_prompt
+    # Check if the message is in Arabic
+    if any(0x0600 <= ord(char) <= 0x06FF for char in message):
+        enhanced_prompt += "\nRespond in Arabic with clear, concise, and accurate information tailored to the user's query."
+    if user and user.additional_info:
+        enhanced_prompt += f"\nUser Profile: {user.additional_info}\nConversation Style: {user.conversation_style or 'default'}"
+    return enhanced_prompt
+
 @router.get("/api/settings")
 async def get_settings(user: User = Depends(current_active_user)):
     if not user:
@@ -166,13 +176,18 @@ async def chat_endpoint(
     # Use user's preferred model if set
     preferred_model = user.preferred_model if user else None
     model_name, api_endpoint = select_model(req.message, input_type="text", preferred_model=preferred_model)
-    system_prompt = req.system_prompt
-    if user and user.additional_info:
-        system_prompt = f"{system_prompt}\nUser Profile: {user.additional_info}\nConversation Style: {user.conversation_style or 'default'}"
+    
+    # Check model availability
+    is_available, api_key, selected_endpoint = check_model_availability(model_name, HF_TOKEN)
+    if not is_available:
+        logger.error(f"Model {model_name} is not available at {api_endpoint}")
+        raise HTTPException(status_code=503, detail=f"Model {model_name} is not available. Please try another model.")
+    
+    system_prompt = enhance_system_prompt(req.system_prompt, req.message, user)
     
     stream = request_generation(
-        api_key=HF_TOKEN,
-        api_base=api_endpoint,
+        api_key=api_key,
+        api_base=selected_endpoint,
         message=req.message,
         system_prompt=system_prompt,
         model_name=model_name,
@@ -183,19 +198,39 @@ async def chat_endpoint(
         input_type="text",
         output_format=req.output_format
     )
+    
     if req.output_format == "audio":
         audio_chunks = []
-        for chunk in stream:
-            if isinstance(chunk, bytes):
-                audio_chunks.append(chunk)
-        audio_data = b"".join(audio_chunks)
-        return StreamingResponse(io.BytesIO(audio_data), media_type="audio/wav")
+        try:
+            for chunk in stream:
+                if isinstance(chunk, bytes):
+                    audio_chunks.append(chunk)
+                else:
+                    logger.warning(f"Unexpected non-bytes chunk in audio stream: {chunk}")
+            if not audio_chunks:
+                logger.error("No audio data generated.")
+                raise HTTPException(status_code=500, detail="No audio data generated.")
+            audio_data = b"".join(audio_chunks)
+            return StreamingResponse(io.BytesIO(audio_data), media_type="audio/wav")
+        except Exception as e:
+            logger.error(f"Audio generation failed: {e}")
+            raise HTTPException(status_code=500, detail=f"Audio generation failed: {str(e)}")
+    
     response_chunks = []
-    for chunk in stream:
-        if isinstance(chunk, str):
-            response_chunks.append(chunk)
-    response = "".join(response_chunks)
-    logger.info(f"Chat response: {response}")
+    try:
+        for chunk in stream:
+            if isinstance(chunk, str):
+                response_chunks.append(chunk)
+            else:
+                logger.warning(f"Unexpected non-string chunk in text stream: {chunk}")
+        response = "".join(response_chunks)
+        if not response.strip():
+            logger.error("Empty response generated.")
+            raise HTTPException(status_code=500, detail="Empty response generated from model.")
+        logger.info(f"Chat response: {response[:100]}...")  # Log first 100 chars
+    except Exception as e:
+        logger.error(f"Chat generation failed: {e}")
+        raise HTTPException(status_code=500, detail=f"Chat generation failed: {str(e)}")
     
     if user and conversation:
         assistant_msg = Message(role="assistant", content=response, conversation_id=conversation.id)
@@ -244,12 +279,19 @@ async def audio_transcription_endpoint(
         db.commit()
     
     model_name, api_endpoint = select_model("transcribe audio", input_type="audio")
+    
+    # Check model availability
+    is_available, api_key, selected_endpoint = check_model_availability(model_name, HF_TOKEN)
+    if not is_available:
+        logger.error(f"Model {model_name} is not available at {api_endpoint}")
+        raise HTTPException(status_code=503, detail=f"Model {model_name} is not available. Please try another model.")
+    
     audio_data = await file.read()
     stream = request_generation(
-        api_key=HF_TOKEN,
-        api_base=api_endpoint,
+        api_key=api_key,
+        api_base=selected_endpoint,
         message="Transcribe audio",
-        system_prompt="Transcribe the provided audio using Whisper.",
+        system_prompt="Transcribe the provided audio using Whisper. Ensure accurate transcription in the detected language.",
         model_name=model_name,
         temperature=0.7,
         max_new_tokens=2048,
@@ -258,11 +300,20 @@ async def audio_transcription_endpoint(
         output_format="text"
     )
     response_chunks = []
-    for chunk in stream:
-        if isinstance(chunk, str):
-            response_chunks.append(chunk)
-    response = "".join(response_chunks)
-    logger.info(f"Audio transcription response: {response}")
+    try:
+        for chunk in stream:
+            if isinstance(chunk, str):
+                response_chunks.append(chunk)
+            else:
+                logger.warning(f"Unexpected non-string chunk in transcription stream: {chunk}")
+        response = "".join(response_chunks)
+        if not response.strip():
+            logger.error("Empty transcription generated.")
+            raise HTTPException(status_code=500, detail="Empty transcription generated from model.")
+        logger.info(f"Audio transcription response: {response[:100]}...")
+    except Exception as e:
+        logger.error(f"Audio transcription failed: {e}")
+        raise HTTPException(status_code=500, detail=f"Audio transcription failed: {str(e)}")
     
     if user and conversation:
         assistant_msg = Message(role="assistant", content=response, conversation_id=conversation.id)
@@ -290,12 +341,22 @@ async def text_to_speech_endpoint(
         await handle_session(request)
     
     text = req.get("text", "")
+    if not text.strip():
+        raise HTTPException(status_code=400, detail="Text input is required for text-to-speech.")
+    
     model_name, api_endpoint = select_model("text to speech", input_type="tts")
+    
+    # Check model availability
+    is_available, api_key, selected_endpoint = check_model_availability(model_name, HF_TOKEN)
+    if not is_available:
+        logger.error(f"Model {model_name} is not available at {api_endpoint}")
+        raise HTTPException(status_code=503, detail=f"Model {model_name} is not available. Please try another model.")
+    
     stream = request_generation(
-        api_key=HF_TOKEN,
-        api_base=api_endpoint,
+        api_key=api_key,
+        api_base=selected_endpoint,
         message=text,
-        system_prompt="Convert the provided text to speech using a text-to-speech model.",
+        system_prompt="Convert the provided text to speech using a text-to-speech model. Ensure clear and natural pronunciation, especially for Arabic text.",
         model_name=model_name,
         temperature=0.7,
         max_new_tokens=2048,
@@ -303,11 +364,20 @@ async def text_to_speech_endpoint(
         output_format="audio"
     )
     audio_chunks = []
-    for chunk in stream:
-        if isinstance(chunk, bytes):
-            audio_chunks.append(chunk)
-    audio_data = b"".join(audio_chunks)
-    return StreamingResponse(io.BytesIO(audio_data), media_type="audio/wav")
+    try:
+        for chunk in stream:
+            if isinstance(chunk, bytes):
+                audio_chunks.append(chunk)
+            else:
+                logger.warning(f"Unexpected non-bytes chunk in TTS stream: {chunk}")
+        if not audio_chunks:
+            logger.error("No audio data generated for TTS.")
+            raise HTTPException(status_code=500, detail="No audio data generated for text-to-speech.")
+        audio_data = b"".join(audio_chunks)
+        return StreamingResponse(io.BytesIO(audio_data), media_type="audio/wav")
+    except Exception as e:
+        logger.error(f"Text-to-speech generation failed: {e}")
+        raise HTTPException(status_code=500, detail=f"Text-to-speech generation failed: {str(e)}")
 
 @router.post("/api/code")
 async def code_endpoint(
@@ -323,16 +393,27 @@ async def code_endpoint(
     task = req.get("task")
     code = req.get("code", "")
     output_format = req.get("output_format", "text")
+    if not task:
+        raise HTTPException(status_code=400, detail="Task description is required.")
+    
     prompt = f"Generate code for task: {task} using {framework}. Existing code: {code}"
     preferred_model = user.preferred_model if user else None
     model_name, api_endpoint = select_model(prompt, input_type="text", preferred_model=preferred_model)
-    system_prompt = "You are a coding expert. Provide detailed, well-commented code with examples and explanations."
-    if user and user.additional_info:
-        system_prompt = f"{system_prompt}\nUser Profile: {user.additional_info}\nConversation Style: {user.conversation_style or 'default'}"
+    
+    # Check model availability
+    is_available, api_key, selected_endpoint = check_model_availability(model_name, HF_TOKEN)
+    if not is_available:
+        logger.error(f"Model {model_name} is not available at {api_endpoint}")
+        raise HTTPException(status_code=503, detail=f"Model {model_name} is not available. Please try another model.")
+    
+    system_prompt = enhance_system_prompt(
+        "You are a coding expert. Provide detailed, well-commented code with examples and explanations.",
+        prompt, user
+    )
     
     stream = request_generation(
-        api_key=HF_TOKEN,
-        api_base=api_endpoint,
+        api_key=api_key,
+        api_base=selected_endpoint,
         message=prompt,
         system_prompt=system_prompt,
         model_name=model_name,
@@ -343,17 +424,36 @@ async def code_endpoint(
     )
     if output_format == "audio":
         audio_chunks = []
-        for chunk in stream:
-            if isinstance(chunk, bytes):
-                audio_chunks.append(chunk)
-        audio_data = b"".join(audio_chunks)
-        return StreamingResponse(io.BytesIO(audio_data), media_type="audio/wav")
+        try:
+            for chunk in stream:
+                if isinstance(chunk, bytes):
+                    audio_chunks.append(chunk)
+                else:
+                    logger.warning(f"Unexpected non-bytes chunk in code audio stream: {chunk}")
+            if not audio_chunks:
+                logger.error("No audio data generated for code.")
+                raise HTTPException(status_code=500, detail="No audio data generated for code.")
+            audio_data = b"".join(audio_chunks)
+            return StreamingResponse(io.BytesIO(audio_data), media_type="audio/wav")
+        except Exception as e:
+            logger.error(f"Code audio generation failed: {e}")
+            raise HTTPException(status_code=500, detail=f"Code audio generation failed: {str(e)}")
+    
     response_chunks = []
-    for chunk in stream:
-        if isinstance(chunk, str):
-            response_chunks.append(chunk)
-    response = "".join(response_chunks)
-    return {"generated_code": response}
+    try:
+        for chunk in stream:
+            if isinstance(chunk, str):
+                response_chunks.append(chunk)
+            else:
+                logger.warning(f"Unexpected non-string chunk in code stream: {chunk}")
+        response = "".join(response_chunks)
+        if not response.strip():
+            logger.error("Empty code response generated.")
+            raise HTTPException(status_code=500, detail="Empty code response generated from model.")
+        return {"generated_code": response}
+    except Exception as e:
+        logger.error(f"Code generation failed: {e}")
+        raise HTTPException(status_code=500, detail=f"Code generation failed: {str(e)}")
 
 @router.post("/api/analysis")
 async def analysis_endpoint(
@@ -367,15 +467,26 @@ async def analysis_endpoint(
     
     message = req.get("text", "")
     output_format = req.get("output_format", "text")
+    if not message.strip():
+        raise HTTPException(status_code=400, detail="Text input is required for analysis.")
+    
     preferred_model = user.preferred_model if user else None
     model_name, api_endpoint = select_model(message, input_type="text", preferred_model=preferred_model)
-    system_prompt = "You are an expert analyst. Provide detailed analysis with step-by-step reasoning and examples."
-    if user and user.additional_info:
-        system_prompt = f"{system_prompt}\nUser Profile: {user.additional_info}\nConversation Style: {user.conversation_style or 'default'}"
+    
+    # Check model availability
+    is_available, api_key, selected_endpoint = check_model_availability(model_name, HF_TOKEN)
+    if not is_available:
+        logger.error(f"Model {model_name} is not available at {api_endpoint}")
+        raise HTTPException(status_code=503, detail=f"Model {model_name} is not available. Please try another model.")
+    
+    system_prompt = enhance_system_prompt(
+        "You are an expert analyst. Provide detailed analysis with step-by-step reasoning and examples.",
+        message, user
+    )
     
     stream = request_generation(
-        api_key=HF_TOKEN,
-        api_base=api_endpoint,
+        api_key=api_key,
+        api_base=selected_endpoint,
         message=message,
         system_prompt=system_prompt,
         model_name=model_name,
@@ -386,17 +497,36 @@ async def analysis_endpoint(
     )
     if output_format == "audio":
         audio_chunks = []
-        for chunk in stream:
-            if isinstance(chunk, bytes):
-                audio_chunks.append(chunk)
-        audio_data = b"".join(audio_chunks)
-        return StreamingResponse(io.BytesIO(audio_data), media_type="audio/wav")
+        try:
+            for chunk in stream:
+                if isinstance(chunk, bytes):
+                    audio_chunks.append(chunk)
+                else:
+                    logger.warning(f"Unexpected non-bytes chunk in analysis audio stream: {chunk}")
+            if not audio_chunks:
+                logger.error("No audio data generated for analysis.")
+                raise HTTPException(status_code=500, detail="No audio data generated for analysis.")
+            audio_data = b"".join(audio_chunks)
+            return StreamingResponse(io.BytesIO(audio_data), media_type="audio/wav")
+        except Exception as e:
+            logger.error(f"Analysis audio generation failed: {e}")
+            raise HTTPException(status_code=500, detail=f"Analysis audio generation failed: {str(e)}")
+    
     response_chunks = []
-    for chunk in stream:
-        if isinstance(chunk, str):
-            response_chunks.append(chunk)
-    response = "".join(response_chunks)
-    return {"analysis": response}
+    try:
+        for chunk in stream:
+            if isinstance(chunk, str):
+                response_chunks.append(chunk)
+            else:
+                logger.warning(f"Unexpected non-string chunk in analysis stream: {chunk}")
+        response = "".join(response_chunks)
+        if not response.strip():
+            logger.error("Empty analysis response generated.")
+            raise HTTPException(status_code=500, detail="Empty analysis response generated from model.")
+        return {"analysis": response}
+    except Exception as e:
+        logger.error(f"Analysis generation failed: {e}")
+        raise HTTPException(status_code=500, detail=f"Analysis generation failed: {str(e)}")
 
 @router.post("/api/image-analysis")
 async def image_analysis_endpoint(
@@ -430,14 +560,22 @@ async def image_analysis_endpoint(
     
     preferred_model = user.preferred_model if user else None
     model_name, api_endpoint = select_model("analyze image", input_type="image", preferred_model=preferred_model)
+    
+    # Check model availability
+    is_available, api_key, selected_endpoint = check_model_availability(model_name, HF_TOKEN)
+    if not is_available:
+        logger.error(f"Model {model_name} is not available at {api_endpoint}")
+        raise HTTPException(status_code=503, detail=f"Model {model_name} is not available. Please try another model.")
+    
     image_data = await file.read()
-    system_prompt = "You are an expert in image analysis. Provide detailed descriptions or classifications based on the query."
-    if user and user.additional_info:
-        system_prompt = f"{system_prompt}\nUser Profile: {user.additional_info}\nConversation Style: {user.conversation_style or 'default'}"
+    system_prompt = enhance_system_prompt(
+        "You are an expert in image analysis. Provide detailed descriptions or classifications based on the query.",
+        "Analyze this image", user
+    )
     
     stream = request_generation(
-        api_key=HF_TOKEN,
-        api_base=api_endpoint,
+        api_key=api_key,
+        api_base=selected_endpoint,
         message="Analyze this image",
         system_prompt=system_prompt,
         model_name=model_name,
@@ -449,36 +587,59 @@ async def image_analysis_endpoint(
     )
     if output_format == "audio":
         audio_chunks = []
-        for chunk in stream:
-            if isinstance(chunk, bytes):
-                audio_chunks.append(chunk)
-        audio_data = b"".join(audio_chunks)
-        return StreamingResponse(io.BytesIO(audio_data), media_type="audio/wav")
-    response_chunks = []
-    for chunk in stream:
-        if isinstance(chunk, str):
-            response_chunks.append(chunk)
-    response = "".join(response_chunks)
-    
-    if user and conversation:
-        assistant_msg = Message(role="assistant", content=response, conversation_id=conversation.id)
-        db.add(assistant_msg)
-        db.commit()
-        conversation.updated_at = datetime.utcnow()
-        db.commit()
-        return {
-            "image_analysis": response,
-            "conversation_id": conversation.conversation_id,
-            "conversation_url": f"https://mgzon-mgzon-app.hf.space/chat/{conversation.conversation_id}",
-            "conversation_title": conversation.title
-        }
+        try:
+            for chunk in stream:
+                if isinstance(chunk, bytes):
+                    audio_chunks.append(chunk)
+                else:
+                    logger.warning(f"Unexpected non-bytes chunk in image analysis audio stream: {chunk}")
+            if not audio_chunks:
+                logger.error("No audio data generated for image analysis.")
+                raise HTTPException(status_code=500, detail="No audio data generated for image analysis.")
+            audio_data = b"".join(audio_chunks)
+            return StreamingResponse(io.BytesIO(audio_data), media_type="audio/wav")
+        except Exception as e:
+            logger.error(f"Image analysis audio generation failed: {e}")
+            raise HTTPException(status_code=500, detail=f"Image analysis audio generation failed: {str(e)}")
     
-    return {"image_analysis": response}
+    response_chunks = []
+    try:
+        for chunk in stream:
+            if isinstance(chunk, str):
+                response_chunks.append(chunk)
+            else:
+                logger.warning(f"Unexpected non-string chunk in image analysis stream: {chunk}")
+        response = "".join(response_chunks)
+        if not response.strip():
+            logger.error("Empty image analysis response generated.")
+            raise HTTPException(status_code=500, detail="Empty image analysis response generated from model.")
+        
+        if user and conversation:
+            assistant_msg = Message(role="assistant", content=response, conversation_id=conversation.id)
+            db.add(assistant_msg)
+            db.commit()
+            conversation.updated_at = datetime.utcnow()
+            db.commit()
+            return {
+                "image_analysis": response,
+                "conversation_id": conversation.conversation_id,
+                "conversation_url": f"https://mgzon-mgzon-app.hf.space/chat/{conversation.conversation_id}",
+                "conversation_title": conversation.title
+            }
+        
+        return {"image_analysis": response}
+    except Exception as e:
+        logger.error(f"Image analysis failed: {e}")
+        raise HTTPException(status_code=500, detail=f"Image analysis failed: {str(e)}")
 
 @router.get("/api/test-model")
 async def test_model(model: str = MODEL_NAME, endpoint: str = ROUTER_API_URL):
     try:
-        _, api_key, selected_endpoint = check_model_availability(model, HF_TOKEN)
+        is_available, api_key, selected_endpoint = check_model_availability(model, HF_TOKEN)
+        if not is_available:
+            logger.error(f"Model {model} is not available at {endpoint}")
+            raise HTTPException(status_code=503, detail=f"Model {model} is not available.")
+        
         client = OpenAI(api_key=api_key, base_url=selected_endpoint, timeout=60.0)
         response = client.chat.completions.create(
             model=model,
@@ -487,7 +648,8 @@ async def test_model(model: str = MODEL_NAME, endpoint: str = ROUTER_API_URL):
         )
         return {"status": "success", "response": response.choices[0].message.content}
     except Exception as e:
-        return {"status": "error", "message": str(e)}
+        logger.error(f"Test model failed: {e}")
+        raise HTTPException(status_code=500, detail=f"Test model failed: {str(e)}")
 
 @router.post("/api/conversations", response_model=ConversationOut)
 async def create_conversation(

static/js/chat.js

@@ -57,6 +57,11 @@ function autoResizeTextarea() {
   }
 }
 
+// Detect Arabic text
+function isArabicText(text) {
+  return /[\u0600-\u06FF]/.test(text);
+}
+
 // Initialize page
 document.addEventListener('DOMContentLoaded', async () => {
   AOS.init({
@@ -98,9 +103,10 @@ function updateSendButtonState() {
   }
 }
 
-// Render markdown content
+// Render markdown content with RTL support
 function renderMarkdown(el) {
   const raw = el.dataset.text || '';
+  const isArabic = isArabicText(raw);
   const html = marked.parse(raw, {
     gfm: true,
     breaks: true,
@@ -108,7 +114,7 @@ function renderMarkdown(el) {
     smartypants: false,
     headerIds: false,
   });
-  el.innerHTML = `<div class="md-content">${html}</div>`;
+  el.innerHTML = `<div class="md-content" style="direction: ${isArabic ? 'rtl' : 'ltr'}; text-align: ${isArabic ? 'right' : 'left'};">${html}</div>`;
   const wrapper = el.querySelector('.md-content');
   wrapper.querySelectorAll('table').forEach(t => {
     if (!t.parentNode.classList?.contains('table-wrapper')) {
@@ -148,7 +154,7 @@ function leaveChatView() {
 // Add chat bubble
 function addMsg(who, text) {
   const div = document.createElement('div');
-  div.className = `bubble ${who === 'user' ? 'bubble-user' : 'bubble-assist'}`;
+  div.className = `bubble ${who === 'user' ? 'bubble-user' : 'bubble-assist'} ${isArabicText(text) ? 'rtl' : ''}`;
   div.dataset.text = text;
   renderMarkdown(div);
   if (uiElements.chatBox) {
@@ -230,7 +236,7 @@ function startVoiceRecording() {
     mediaRecorder.addEventListener('dataavailable', event => audioChunks.push(event.data));
   }).catch(err => {
     console.error('Error accessing microphone:', err);
-    alert('Failed to access microphone. Please check permissions.');
+    alert('فشل الوصول إلى الميكروفون. من فضلك، تحقق من الأذونات.');
     isRecording = false;
     if (uiElements.sendBtn) uiElements.sendBtn.classList.remove('recording');
   });
@@ -253,8 +259,8 @@ function stopVoiceRecording() {
 // Send audio message
 async function submitAudioMessage(formData) {
   enterChatView();
-  addMsg('user', 'Voice message');
-  conversationHistory.push({ role: 'user', content: 'Voice message' });
+  addMsg('user', 'رسالة صوتية');
+  conversationHistory.push({ role: 'user', content: 'رسالة صوتية' });
   sessionStorage.setItem('conversationHistory', JSON.stringify(conversationHistory));
   streamMsg = addMsg('assistant', '');
   const loadingEl = document.createElement('span');
@@ -267,9 +273,12 @@ async function submitAudioMessage(formData) {
 
   try {
     const response = await sendRequest('/api/audio-transcription', formData);
+    if (!response.ok) {
+      throw new Error(`Request failed with status ${response.status}`);
+    }
     const data = await response.json();
-    if (!data.transcription) throw new Error('No transcription received from server');
-    const transcription = data.transcription || 'Error: No transcription generated.';
+    if (!data.transcription) throw new Error('لم يتم استلام نص النسخ من الخادم');
+    const transcription = data.transcription || 'خطأ: لم يتم إنشاء نص النسخ.';
     if (streamMsg) {
       streamMsg.dataset.text = transcription;
       renderMarkdown(streamMsg);
@@ -282,7 +291,7 @@ async function submitAudioMessage(formData) {
     }
     if (checkAuth() && data.conversation_id) {
       currentConversationId = data.conversation_id;
-      currentConversationTitle = data.conversation_title || 'Untitled Conversation';
+      currentConversationTitle = data.conversation_title || 'محادثة بدون عنوان';
       if (uiElements.conversationTitle) uiElements.conversationTitle.textContent = currentConversationTitle;
       history.pushState(null, '', `/chat/${currentConversationId}`);
       await loadConversations();
@@ -297,12 +306,35 @@ async function submitAudioMessage(formData) {
 async function sendRequest(endpoint, body, headers = {}) {
   const token = checkAuth();
   if (token) headers['Authorization'] = `Bearer ${token}`;
-  return await fetch(endpoint, {
-    method: 'POST',
-    body,
-    headers,
-    signal: abortController.signal,
-  });
+  try {
+    const response = await fetch(endpoint, {
+      method: 'POST',
+      body,
+      headers,
+      signal: abortController?.signal,
+    });
+    if (!response.ok) {
+      if (response.status === 403) {
+        if (uiElements.messageLimitWarning) uiElements.messageLimitWarning.classList.remove('hidden');
+        throw new Error('تم الوصول إلى الحد الأقصى للرسائل. من فضلك، سجل الدخول للمتابعة.');
+      }
+      if (response.status === 401) {
+        localStorage.removeItem('token');
+        window.location.href = '/login';
+        throw new Error('غير مصرح. من فضلك، سجل الدخول مرة أخرى.');
+      }
+      if (response.status === 503) {
+        throw new Error('النموذج غير متاح حاليًا. من فضلك، حاول استخدام نموذج آخر.');
+      }
+      throw new Error(`فشل الطلب: ${response.status}`);
+    }
+    return response;
+  } catch (error) {
+    if (error.name === 'AbortError') {
+      throw new Error('تم إلغاء الطلب');
+    }
+    throw error;
+  }
 }
 
 // Helper to update UI during request
@@ -329,12 +361,13 @@ function finalizeRequest() {
 function handleRequestError(error) {
   if (streamMsg) {
     streamMsg.querySelector('.loading')?.remove();
-    streamMsg.dataset.text = `Error: ${error.message || 'An error occurred during the request.'}`;
+    streamMsg.dataset.text = `خطأ: ${error.message || 'حدث خطأ أثناء الطلب.'}`;
     renderMarkdown(streamMsg);
     streamMsg.dataset.done = '1';
     streamMsg = null;
   }
-  console.error('Request error:', error);
+  console.error('خطأ في الطلب:', error);
+  alert(`خطأ: ${error.message || 'حدث خطأ أثناء الطلب.'}`);
   isRequestActive = false;
   abortController = null;
   if (uiElements.sendBtn) uiElements.sendBtn.style.display = 'inline-flex';
@@ -348,7 +381,7 @@ async function loadConversations() {
     const response = await fetch('/api/conversations', {
       headers: { 'Authorization': `Bearer ${checkAuth()}` }
     });
-    if (!response.ok) throw new Error('Failed to load conversations');
+    if (!response.ok) throw new Error('فشل تحميل المحادثات');
     const conversations = await response.json();
     if (uiElements.conversationList) {
       uiElements.conversationList.innerHTML = '';
@@ -357,13 +390,13 @@ async function loadConversations() {
         li.className = `flex items-center justify-between text-white hover:bg-gray-700 p-2 rounded cursor-pointer transition-colors ${conv.conversation_id === currentConversationId ? 'bg-gray-700' : ''}`;
         li.dataset.conversationId = conv.conversation_id;
         li.innerHTML = `
-          <div class="flex items-center flex-1" data-conversation-id="${conv.conversation_id}">
+          <div class="flex items-center flex-1" style="direction: ${isArabicText(conv.title) ? 'rtl' : 'ltr'};" data-conversation-id="${conv.conversation_id}">
             <svg class="w-5 h-5 mr-2" fill="none" stroke="currentColor" viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg">
               <path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M8 10h.01M12 10h.01M16 10h.01M9 16H5a2 2 0 01-2-2V6a2 2 0 012-2h14a2 2 0 012 2v8a2 2 0 01-2 2h-5l-5 5v-5z"></path>
             </svg>
-            <span class="truncate flex-1">${conv.title || 'Untitled Conversation'}</span>
+            <span class="truncate flex-1">${conv.title || 'محادثة بدون عنوان'}</span>
           </div>
-          <button class="delete-conversation-btn text-red-400 hover:text-red-600 p-1" title="Delete Conversation" data-conversation-id="${conv.conversation_id}">
+          <button class="delete-conversation-btn text-red-400 hover:text-red-600 p-1" title="حذف المحادثة" data-conversation-id="${conv.conversation_id}">
             <svg class="w-5 h-5" fill="none" stroke="currentColor" viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg">
               <path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M19 7l-.867 12.142A2 2 0 0116.138 21H7.862a2 2 0 01-1.995-1.858L5 7m5-4h4M3 7h18"></path>
             </svg>
@@ -375,7 +408,8 @@ async function loadConversations() {
       });
     }
   } catch (error) {
-    console.error('Error loading conversations:', error);
+    console.error('خطأ في تحميل المحادثات:', error);
+    alert('فشل تحميل المحادثات. من فضلك، حاول مرة أخرى.');
   }
 }
 
@@ -387,11 +421,11 @@ async function loadConversation(conversationId) {
     });
     if (!response.ok) {
       if (response.status === 401) window.location.href = '/login';
-      throw new Error('Failed to load conversation');
+      throw new Error('فشل تحميل المحادثة');
     }
     const data = await response.json();
     currentConversationId = data.conversation_id;
-    currentConversationTitle = data.title || 'Untitled Conversation';
+    currentConversationTitle = data.title || 'محادثة بدون عنوان';
     conversationHistory = data.messages.map(msg => ({ role: msg.role, content: msg.content }));
     if (uiElements.chatBox) uiElements.chatBox.innerHTML = '';
     conversationHistory.forEach(msg => addMsg(msg.role, msg.content));
@@ -400,14 +434,14 @@ async function loadConversation(conversationId) {
     history.pushState(null, '', `/chat/${currentConversationId}`);
     toggleSidebar(false);
   } catch (error) {
-    console.error('Error loading conversation:', error);
-    alert('Failed to load conversation. Please try again or log in.');
+    console.error('خطأ في تحميل المحادثة:', error);
+    alert('فشل تحميل المحادثة. من فضلك، حاول مرة أخرى أو سجل الدخول.');
   }
 }
 
 // Delete conversation
 async function deleteConversation(conversationId) {
-  if (!confirm('Are you sure you want to delete this conversation?')) return;
+  if (!confirm('هل أنت متأكد من حذف هذه المحادثة؟')) return;
   try {
     const response = await fetch(`/api/conversations/${conversationId}`, {
       method: 'DELETE',
@@ -415,7 +449,7 @@ async function deleteConversation(conversationId) {
     });
     if (!response.ok) {
       if (response.status === 401) window.location.href = '/login';
-      throw new Error('Failed to delete conversation');
+      throw new Error('فشل حذف المحادثة');
     }
     if (conversationId === currentConversationId) {
       clearAllMessages();
@@ -425,8 +459,8 @@ async function deleteConversation(conversationId) {
     }
     await loadConversations();
   } catch (error) {
-    console.error('Error deleting conversation:', error);
-    alert('Failed to delete conversation. Please try again.');
+    console.error('خطأ في حذف المحادثة:', error);
+    alert('فشل حذف المحادثة. من فضلك، حاول مرة أخرى.');
   }
 }
 
@@ -441,16 +475,16 @@ async function saveMessageToConversation(conversationId, role, content) {
       },
       body: JSON.stringify({ role, content })
     });
-    if (!response.ok) throw new Error('Failed to save message');
+    if (!response.ok) throw new Error('فشل حفظ الرسالة');
   } catch (error) {
-    console.error('Error saving message:', error);
+    console.error('خطأ في حفظ الرسالة:', error);
   }
 }
 
 // Create new conversation
 async function createNewConversation() {
   if (!checkAuth()) {
-    alert('Please log in to create a new conversation.');
+    alert('من فضلك، سجل الدخول لإنشاء محادثة جديدة.');
     window.location.href = '/login';
     return;
   }
@@ -461,14 +495,14 @@ async function createNewConversation() {
         'Content-Type': 'application/json',
         'Authorization': `Bearer ${checkAuth()}`
       },
-      body: JSON.stringify({ title: 'New Conversation' })
+      body: JSON.stringify({ title: 'محادثة جديدة' })
     });
     if (!response.ok) {
       if (response.status === 401) {
         localStorage.removeItem('token');
         window.location.href = '/login';
       }
-      throw new Error('Failed to create conversation');
+      throw new Error('فشل إنشاء المحادثة');
     }
     const data = await response.json();
     currentConversationId = data.conversation_id;
@@ -482,8 +516,8 @@ async function createNewConversation() {
     await loadConversations();
     toggleSidebar(false);
   } catch (error) {
-    console.error('Error creating conversation:', error);
-    alert('Failed to create new conversation. Please try again.');
+    console.error('خطأ في إنشاء المحادثة:', error);
+    alert('فشل إنشاء محادثة جديدة. من فضلك، حاول مرة أخرى.');
   }
 }
 
@@ -498,14 +532,14 @@ async function updateConversationTitle(conversationId, newTitle) {
       },
       body: JSON.stringify({ title: newTitle })
     });
-    if (!response.ok) throw new Error('Failed to update title');
+    if (!response.ok) throw new Error('فشل تحديث العنوان');
     const data = await response.json();
     currentConversationTitle = data.title;
     if (uiElements.conversationTitle) uiElements.conversationTitle.textContent = currentConversationTitle;
     await loadConversations();
   } catch (error) {
-    console.error('Error updating title:', error);
-    alert('Failed to update conversation title.');
+    console.error('خطأ في تحديث العنوان:', error);
+    alert('فشل تحديث عنوان المحادثة.');
   }
 }
 
@@ -598,7 +632,7 @@ async function submitMessage() {
     if (file.type.startsWith('image/')) {
       endpoint = '/api/image-analysis';
       inputType = 'image';
-      message = 'Analyze this image';
+      message = 'تحليل هذه الصورة';
       formData = new FormData();
       formData.append('file', file);
       formData.append('output_format', 'text');
@@ -608,14 +642,16 @@ async function submitMessage() {
     if (file.type.startsWith('audio/')) {
       endpoint = '/api/audio-transcription';
       inputType = 'audio';
-      message = 'Transcribe this audio';
+      message = 'نسخ هذا الصوت';
       formData = new FormData();
       formData.append('file', file);
     }
   } else if (message) {
     payload = {
       message,
-      system_prompt: 'You are an expert assistant providing detailed, comprehensive, and well-structured responses.',
+      system_prompt: isArabicText(message)
+        ? 'أنت مساعد ذكي تقدم إجابات مفصلة ومنظمة باللغة العربية، مع ضمان الدقة والوضوح.'
+        : 'You are an expert assistant providing detailed, comprehensive, and well-structured responses.',
       history: conversationHistory,
       temperature: 0.7,
       max_new_tokens: 128000,
@@ -643,31 +679,10 @@ async function submitMessage() {
 
   try {
     const response = await sendRequest(endpoint, payload ? JSON.stringify(payload) : formData, headers);
-    if (!response.ok) {
-      if (response.status === 403) {
-        if (uiElements.messageLimitWarning) uiElements.messageLimitWarning.classList.remove('hidden');
-        if (uiElements.input) uiElements.input.disabled = true;
-        if (streamMsg) streamMsg.querySelector('.loading')?.remove();
-        streamMsg = null;
-        isRequestActive = false;
-        abortController = null;
-        if (uiElements.sendBtn) uiElements.sendBtn.style.display = 'inline-flex';
-        if (uiElements.stopBtn) uiElements.stopBtn.style.display = 'none';
-        setTimeout(() => window.location.href = '/login', 3000);
-        return;
-      }
-      if (response.status === 401) {
-        localStorage.removeItem('token');
-        window.location.href = '/login';
-        return;
-      }
-      throw new Error(`Request failed with status ${response.status}`);
-    }
-
     if (endpoint === '/api/audio-transcription') {
       const data = await response.json();
-      if (!data.transcription) throw new Error('No transcription received from server');
-      const transcription = data.transcription || 'Error: No transcription generated.';
+      if (!data.transcription) throw new Error('لم يتم استلام نص النسخ من الخادم');
+      const transcription = data.transcription || 'خطأ: لم يتم إنشاء نص النسخ.';
       if (streamMsg) {
         streamMsg.dataset.text = transcription;
         renderMarkdown(streamMsg);
@@ -680,14 +695,14 @@ async function submitMessage() {
       }
       if (checkAuth() && data.conversation_id) {
         currentConversationId = data.conversation_id;
-        currentConversationTitle = data.conversation_title || 'Untitled Conversation';
+        currentConversationTitle = data.conversation_title || 'محادثة بدون عنوان';
         if (uiElements.conversationTitle) uiElements.conversationTitle.textContent = currentConversationTitle;
         history.pushState(null, '', `/chat/${currentConversationId}`);
         await loadConversations();
       }
     } else if (endpoint === '/api/image-analysis') {
       const data = await response.json();
-      const analysis = data.image_analysis || 'Error: No analysis generated.';
+      const analysis = data.image_analysis || 'خطأ: لم يتم إنشاء تحليل.';
       if (streamMsg) {
         streamMsg.dataset.text = analysis;
         renderMarkdown(streamMsg);
@@ -700,7 +715,7 @@ async function submitMessage() {
       }
       if (checkAuth() && data.conversation_id) {
         currentConversationId = data.conversation_id;
-        currentConversationTitle = data.conversation_title || 'Untitled Conversation';
+        currentConversationTitle = data.conversation_title || 'محادثة بدون عنوان';
         if (uiElements.conversationTitle) uiElements.conversationTitle.textContent = currentConversationTitle;
         history.pushState(null, '', `/chat/${currentConversationId}`);
         await loadConversations();
@@ -709,7 +724,7 @@ async function submitMessage() {
       const contentType = response.headers.get('Content-Type');
       if (contentType?.includes('application/json')) {
         const data = await response.json();
-        const responseText = data.response || 'Error: No response generated.';
+        const responseText = data.response || 'خطأ: لم يتم إنشاء رد.';
         if (streamMsg) {
           streamMsg.dataset.text = responseText;
           renderMarkdown(streamMsg);
@@ -722,7 +737,7 @@ async function submitMessage() {
         }
         if (checkAuth() && data.conversation_id) {
           currentConversationId = data.conversation_id;
-          currentConversationTitle = data.conversation_title || 'Untitled Conversation';
+          currentConversationTitle = data.conversation_title || 'محادثة بدون عنوان';
           if (uiElements.conversationTitle) uiElements.conversationTitle.textContent = currentConversationTitle;
           history.pushState(null, '', `/chat/${currentConversationId}`);
           await loadConversations();
@@ -733,7 +748,10 @@ async function submitMessage() {
         let buffer = '';
         while (true) {
           const { done, value } = await reader.read();
-          if (done) break;
+          if (done) {
+            if (!buffer.trim()) throw new Error('الرد فارغ من الخادم');
+            break;
+          }
           buffer += decoder.decode(value, { stream: true });
           if (streamMsg) {
             streamMsg.dataset.text = buffer;
@@ -782,12 +800,12 @@ function stopStream(forceCancel = false) {
 if (uiElements.settingsBtn) {
   uiElements.settingsBtn.addEventListener('click', () => {
     if (!checkAuth()) {
-      alert('Please log in to access settings.');
+      alert('من فضلك، سجل الدخول للوصول إلى الإعدادات.');
       window.location.href = '/login';
       return;
     }
     uiElements.settingsModal.classList.remove('hidden');
-    fetch('/api/settings', {  // تعديل: استخدام /api/settings بدل /users/me
+    fetch('/api/settings', {
       headers: { 'Authorization': `Bearer ${checkAuth()}` }
     })
       .then(res => {
@@ -796,12 +814,11 @@ if (uiElements.settingsBtn) {
             localStorage.removeItem('token');
             window.location.href = '/login';
           }
-          throw new Error('Failed to fetch settings');
+          throw new Error('فشل جلب الإعدادات');
         }
         return res.json();
       })
       .then(data => {
-        // تعبئة حقول الـ form
         document.getElementById('display_name').value = data.user_settings.display_name || '';
         document.getElementById('preferred_model').value = data.user_settings.preferred_model || 'standard';
         document.getElementById('job_title').value = data.user_settings.job_title || '';
@@ -810,9 +827,8 @@ if (uiElements.settingsBtn) {
         document.getElementById('additional_info').value = data.user_settings.additional_info || '';
         document.getElementById('conversation_style').value = data.user_settings.conversation_style || 'default';
 
-        // تعبئة خيارات preferred_model ديناميكيًا
         const modelSelect = document.getElementById('preferred_model');
-        modelSelect.innerHTML = ''; // تفريغ الخيارات القديمة
+        modelSelect.innerHTML = '';
         data.available_models.forEach(model => {
           const option = document.createElement('option');
           option.value = model.alias;
@@ -820,19 +836,18 @@ if (uiElements.settingsBtn) {
           modelSelect.appendChild(option);
         });
 
-        // تعبئة خيارات conversation_style ديناميكيًا
         const styleSelect = document.getElementById('conversation_style');
-        styleSelect.innerHTML = ''; // تفريغ الخيارات القديمة
+        styleSelect.innerHTML = '';
         data.conversation_styles.forEach(style => {
           const option = document.createElement('option');
           option.value = style;
-          option.textContent = style.charAt(0).toUpperCase() + style.slice(1); // تحويل الحرف الأول لكابيتال
+          option.textContent = style.charAt(0).toUpperCase() + style.slice(1);
           styleSelect.appendChild(option);
         });
       })
       .catch(err => {
-        console.error('Error fetching settings:', err);
-        alert('Failed to load settings. Please try again.');
+        console.error('خطأ في جلب الإعدادات:', err);
+        alert('فشل تحميل الإعدادات. من فضلك، حاول مرة أخرى.');
       });
   });
 }
@@ -847,7 +862,7 @@ if (uiElements.settingsForm) {
   uiElements.settingsForm.addEventListener('submit', (e) => {
     e.preventDefault();
     if (!checkAuth()) {
-      alert('Please log in to save settings.');
+      alert('من فضلك، سجل الدخول لحفظ الإعدادات.');
       window.location.href = '/login';
       return;
     }
@@ -867,18 +882,18 @@ if (uiElements.settingsForm) {
             localStorage.removeItem('token');
             window.location.href = '/login';
           }
-          throw new Error('Failed to update settings');
+          throw new Error('فشل تحديث الإعدادات');
         }
         return res.json();
       })
       .then(() => {
-        alert('Settings updated successfully!');
+        alert('تم تحديث الإعدادات بنجاح!');
         uiElements.settingsModal.classList.add('hidden');
         toggleSidebar(false);
       })
       .catch(err => {
-        console.error('Error updating settings:', err);
-        alert('Error updating settings: ' + err.message);
+        console.error('خطأ في تحديث الإعدادات:', err);
+        alert('خطأ في تحديث الإعدادات: ' + err.message);
       });
   });
 }
@@ -891,10 +906,10 @@ if (uiElements.historyToggle) {
       uiElements.historyToggle.innerHTML = uiElements.conversationList.classList.contains('hidden')
         ? `<svg class="w-5 h-5 mr-2" fill="none" stroke="currentColor" viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg">
              <path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M12 8v4l3 3m6-3a9 9 0 11-18 0 9 9 0 0118 0z"></path>
-           </svg>Show History`
+           </svg>إظهار السجل`
         : `<svg class="w-5 h-5 mr-2" fill="none" stroke="currentColor" viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg">
              <path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M6 18L18 6M6 6l12 12"></path>
-           </svg>Hide History`;
+           </svg>إخفاء السجل`;
     }
   });
 }
@@ -970,8 +985,8 @@ if (uiElements.clearBtn) uiElements.clearBtn.addEventListener('click', clearAllM
 
 if (uiElements.conversationTitle) {
   uiElements.conversationTitle.addEventListener('click', () => {
-    if (!checkAuth()) return alert('Please log in to edit the conversation title.');
-    const newTitle = prompt('Enter new conversation title:', currentConversationTitle || '');
+    if (!checkAuth()) return alert('من فضلك، سجل الدخول لتعديل عنوان المحادثة.');
+    const newTitle = prompt('أدخل عنوان المحادثة الجديد:', currentConversationTitle || '');
     if (newTitle && currentConversationId) {
       updateConversationTitle(currentConversationId, newTitle);
     }

utils/generation.py

@@ -25,10 +25,10 @@ cache = TTLCache(maxsize=int(os.getenv("QUEUE_SIZE", 100)), ttl=600)
 
 # تعريف LATEX_DELIMS
 LATEX_DELIMS = [
-    {"left": "$$  ", "right": "  $$", "display": True},
+    {"left": "$$", "right": "$$", "display": True},
     {"left": "$", "right": "$", "display": False},
-    {"left": "\$$ ", "right": "\ $$", "display": True},
-    {"left": "\$$ ", "right": "\ $$", "display": False},
+    {"left": "\\[", "right": "\\]", "display": True},
+    {"left": "\\(", "right": "\\)", "display": False},
 ]
 
 # إعداد العميل لـ Hugging Face Router API
@@ -37,7 +37,7 @@ BACKUP_HF_TOKEN = os.getenv("BACKUP_HF_TOKEN")
 ROUTER_API_URL = os.getenv("ROUTER_API_URL", "https://router.huggingface.co")
 API_ENDPOINT = os.getenv("API_ENDPOINT", "https://api-inference.huggingface.co")
 FALLBACK_API_ENDPOINT = os.getenv("FALLBACK_API_ENDPOINT", "https://api-inference.huggingface.co")
-MODEL_NAME = os.getenv("MODEL_NAME", "openai/gpt-oss-120b")  # بدون :cerebras
+MODEL_NAME = os.getenv("MODEL_NAME", "openai/gpt-oss-120b")  # النموذج الرئيسي
 SECONDARY_MODEL_NAME = os.getenv("SECONDARY_MODEL_NAME", "mistralai/Mixtral-8x7B-Instruct-v0.1")
 TERTIARY_MODEL_NAME = os.getenv("TERTIARY_MODEL_NAME", "Qwen/Qwen2.5-0.5B-Instruct")
 CLIP_BASE_MODEL = os.getenv("CLIP_BASE_MODEL", "Salesforce/blip-image-captioning-large")
@@ -45,9 +45,8 @@ CLIP_LARGE_MODEL = os.getenv("CLIP_LARGE_MODEL", "openai/clip-vit-large-patch14"
 ASR_MODEL = os.getenv("ASR_MODEL", "openai/whisper-large-v3")
 TTS_MODEL = os.getenv("TTS_MODEL", "facebook/mms-tts-ara")
 
-# Provider endpoints (based on Router API providers)
+# Provider endpoints (بدون together)
 PROVIDER_ENDPOINTS = {
-
     "fireworks-ai": "https://api.fireworks.ai/inference/v1",
     "nebius": "https://api.nebius.ai/v1",
     "novita": "https://api.novita.ai/v1",
@@ -78,7 +77,13 @@ def check_model_availability(model_name: str, api_key: str) -> tuple[bool, str,
         if response.status_code == 200:
             data = response.json().get("data", {})
             providers = data.get("providers", [])
-            # Select the first available provider (e.g., 'cerebras')
+            # Prefer "cerebras" if available
+            for provider in providers:
+                if provider.get("provider") == "cerebras" and provider.get("status") == "live":
+                    endpoint = PROVIDER_ENDPOINTS.get("cerebras", API_ENDPOINT)
+                    logger.info(f"Model {model_name} is available via preferred provider cerebras at {endpoint}")
+                    return True, api_key, endpoint
+            # Fallback to first live provider if cerebras not available
             for provider in providers:
                 if provider.get("status") == "live":
                     provider_name = provider.get("provider")
@@ -410,7 +415,7 @@ def request_generation(
     except Exception as e:
         logger.exception(f"[Gateway] Streaming failed for model {model_name}: {e}")
         if selected_api_key != BACKUP_HF_TOKEN and BACKUP_HF_TOKEN:
-            logger.warning(f"Retrying with backup token for model {model_name}")
+            logger.warning(f"Retrying with backup token for {model_name}")
             for chunk in request_generation(
                 api_key=BACKUP_HF_TOKEN,
                 api_base=selected_endpoint,
@@ -475,27 +480,27 @@ def request_generation(
                             buffer = ""
                         continue
 
-                    if chunk.choices[0].finish_reason in ("stop", "error", "length"):
-                        if buffer:
-                            cached_chunks.append(buffer)
-                            yield buffer
-                            buffer = ""
+                        if chunk.choices[0].finish_reason in ("stop", "error", "length"):
+                            if buffer:
+                                cached_chunks.append(buffer)
+                                yield buffer
+                                buffer = ""
 
-                        if reasoning_started and not reasoning_closed:
-                            cached_chunks.append("assistantfinal")
-                            yield "assistantfinal"
-                            reasoning_closed = True
-
-                        if not saw_visible_output:
-                            cached_chunks.append("No visible output produced.")
-                            yield "No visible output produced."
-                        if chunk.choices[0].finish_reason == "error":
-                            cached_chunks.append(f"Error: Unknown error with fallback model {fallback_model}")
-                            yield f"Error: Unknown error with fallback model {fallback_model}"
-                        elif chunk.choices[0].finish_reason == "length":
-                            cached_chunks.append("Response truncated due to token limit. Please refine your query or request continuation.")
-                            yield "Response truncated due to token limit. Please refine your query or request continuation."
-                        break
+                            if reasoning_started and not reasoning_closed:
+                                cached_chunks.append("assistantfinal")
+                                yield "assistantfinal"
+                                reasoning_closed = True
+
+                            if not saw_visible_output:
+                                cached_chunks.append("No visible output produced.")
+                                yield "No visible output produced."
+                            if chunk.choices[0].finish_reason == "error":
+                                cached_chunks.append(f"Error: Unknown error with fallback model {fallback_model}")
+                                yield f"Error: Unknown error with fallback model {fallback_model}"
+                            elif chunk.choices[0].finish_reason == "length":
+                                cached_chunks.append("Response truncated due to token limit. Please refine your query or request continuation.")
+                                yield "Response truncated due to token limit. Please refine your query or request continuation."
+                            break
 
                 if buffer and output_format == "audio":
                     try:

utils/web_search.py

@@ -1,7 +1,9 @@
+#web_search.py
 import os
 import requests
 from bs4 import BeautifulSoup
 import logging
+import time  # لإضافة التأخير
 
 logger = logging.getLogger(__name__)
 
@@ -23,7 +25,8 @@ def web_search(query: str) -> str:
             snippet = item.get("snippet", "")
             link = item.get("link", "")
             try:
-                page_response = requests.get(link, timeout=5)
+                time.sleep(2)  # إضافة تأخير 2 ثواني بين كل طلب
+                page_response = requests.get(link, timeout=10)
                 page_response.raise_for_status()
                 soup = BeautifulSoup(page_response.text, "html.parser")
                 paragraphs = soup.find_all("p")
@@ -36,3 +39,4 @@ def web_search(query: str) -> str:
     except Exception as e:
         logger.exception("Web search failed")
         return f"Web search error: {e}"
+